Device for controlling an adaptive headlight, adaptive headlight, and motor vehicle

Abstract

The present invention relates to a device (100) for controlling an adaptive headlight (12) for a motor vehicle (10). The device (100) comprises an electronic computing unit (110) configured to receive image data (112) from at least one image acquisition device (16) of the motor vehicle (10), wherein the image data (112) comprise at least a section in front of the motor vehicle (10). Furthermore, the electronic computing unit (110) is configured to feed the image data (112) into a transformer-based image analysis model implemented on the motor vehicle side, which is configured to determine a contextualized driving scene description based on the image data (112). The electronic computing unit (110) is further configured to generate at least one control signal (114) based on the contextualized driving scene description, which is configured to control the adaptive headlight (12, 14).

Description

[0001] The present invention relates to a device for controlling an adaptive headlight. The present invention also relates to an adaptive headlight with such a device. Furthermore, the present invention relates to a motor vehicle with such a device and / or such an adaptive headlight.

[0002] In the field of vehicle lighting, adaptive headlight systems are used that can adjust to current driving conditions. These systems are designed to improve the driver's visibility while minimizing glare for other road users. Common systems typically include sensors and image capture devices that collect various data, as well as control units that process this data and make corresponding adjustments to the headlights. These systems are designed to dynamically control the vehicle's light distribution to ensure optimal road illumination.

[0003] Conventional systems use image data from image acquisition devices to gather information about the vehicle's surroundings. This information is then analyzed by various algorithms to make decisions about adjusting the headlights. Some systems use simple image processing algorithms to detect obstacles or other vehicles and adjust the light distribution accordingly. Other systems use more complex approaches, such as neural networks, to enable a more precise analysis of the driving situation.

[0004] For example, CN 119348536 A describes a method and a system for assisted control of vehicle lighting using a ChatGPT-based language model running on cloud computing resources.

[0005] Despite these advances, challenges still exist in accurately detecting and distinguishing oncoming traffic and other light sources, which can lead to suboptimal control of the headlights.

[0006] The object of the invention is therefore to create the most robust possible means of detecting and distinguishing oncoming traffic and other light sources in order to control an adaptive front light of a motor vehicle.

[0007] This problem is solved by the subject matter of the independent claims. Further advantageous embodiments are specified in the dependent claims and the following description.

[0008] According to a first aspect, a device for controlling an adaptive headlight for a motor vehicle is provided. The device comprises at least one electronic computing unit. The electronic computing unit is configured to receive image data from at least one image acquisition device of the motor vehicle. The image data includes at least a section in front of the motor vehicle. Furthermore, the electronic computing unit is configured to feed the image data into a transformer-based image analysis model implemented on the motor vehicle side. The image analysis model is configured to determine a contextualized driving scene description based on the image data. Additionally, the electronic computing unit is configured to generate at least one control signal based on the contextualized driving scene description. This control signal is configured to control the adaptive headlight.

[0009] One advantage of the device is its reliable, and especially early, dimming of the lights, which largely prevents dazzling oncoming drivers. Another advantage is the avoidance of unnecessary dimming by correctly identifying reflective objects and non-vehicle-related light sources. The use of a transformer-based image analysis model, i.e., a Vision Transformer (ViTs), enables more accurate and faster processing of image data compared to conventional methods such as Convolutional Neural Networks (CNNs). ViTs are capable of analyzing complex driving scenes and taking into account various environmental conditions such as fog, rain, or snow, resulting in more robust and reliable control of the adaptive headlights.This improves driver safety and comfort, minimizing the risk of accidents or distractions caused by incorrectly dimmed headlights, especially high beams. The device processes data locally, i.e., on the vehicle itself or onboard, without the delay of data transmission to a cloud. This allows the system to react quickly to traffic events, potentially even in real time. The vehicle-side version also does not require mobile or Wi-Fi access, as the device functions offline. This ensures that the device remains operational even in tunnels, remote areas, or during network outages. Regarding data security, the image data remains within the vehicle and is not transmitted to external servers or a cloud.

[0010] As used herein, an electronic computing unit is a hardware-based processing unit designed to perform data processing tasks related to vehicle functionality. It can perform a variety of control, analysis, and decision-making processes, particularly in the areas of sensor processing, actuator control, communication, driver assistance, or comfort functions. For example, depending on its architecture, the electronic computing unit may include at least one of the following components: a data processing circuit, a CPU (Central Processing Unit), an ECU (Electronic Control Unit), a SoC (System-on-Chip), an NPU (Neural Processing Unit), a GPU (Graphics Processing Unit), a TPU (Tensor Processing Unit), an MCU (Microcontroller Unit), an FPGA (Field Programmable Gate Array), a DPU (Data Processing Unit), and the like.The electronic computing unit can be coupled or connected to at least one image acquisition device of the vehicle, for example via a communication network in the vehicle, e.g., CAN, Ethernet, or similar. Furthermore, the electronic computing unit can be coupled or connected to the adaptive headlight and / or its control unit.

[0011] Image data can be understood as digital information captured by an image capture device, e.g., camera, sensor, etc., and representing visual content. This can be one or more individual images, video data, etc.

[0012] The at least one image acquisition device of the motor vehicle can be, for example, a camera, a sensor, or the like. This device can be coupled or connected to the electronic computing unit. The at least one image acquisition device can be configured to capture image data of the vehicle's surroundings and to provide this image data.

[0013] As used here, the Transformer-based image analysis model can be understood as a neural network that processes image data using a Transformer core architecture based on self-attention mechanisms. For example, an input image can be divided into a multitude of sub-areas (so-called "patches"), with each patch being assigned a numerical feature vector. The resulting patch vectors are treated as a sequence and fed, along with positional information, to one or more Transformer encoder blocks. Through the weighted aggregation of the patch information within the framework of the attention mechanism, a contextualized representation of the entire image content is created, which can be used for various subsequent applications, such as classification, segmentation, or object recognition.In other words, this can be understood as an image analysis model that uses a transformer core architecture with self-attention mechanisms for the sequential processing of image-based patch inputs to generate a semantically weighted representation of the image content. The contextualized driving scene description can be understood as a detailed and interpreted representation of the environment and / or the driving situation derived from the image data.

[0014] The control signal is an electronic signal used to control a specific function or device, equipment, etc., in this case the adaptive front light.

[0015] Adaptive headlights for motor vehicles can be understood as a lighting system, including intelligent systems, that automatically adjusts the light distribution of a headlight to current driving and environmental conditions in order to optimize visibility for the driver while minimizing glare for other road users. The system takes into account parameters such as vehicle speed, steering angle, road curvature, weather conditions, and the position of oncoming or preceding vehicles. In particular, the high beam can be controlled adaptively, remaining permanently activated while selectively creating shaded areas to avoid dazzling other road users.Adaptive headlights can be implemented in various technical designs, for example, through a segmented headlight unit where individual light segments can be activated or deactivated independently. Other examples include pixel light systems with digital micromirrors (DMD / DLP) for high-resolution light distribution, liquid crystal modulators (LCD shutters) in the light path for dynamic light masking, mechanically adjustable optics (e.g., swiveling lenses or reflectors), and laser-based high-beam systems with adaptive alignment. These diverse technologies enable flexible, context-dependent illumination of the driving area, thus making a significant contribution to vehicle safety and driving comfort.

[0016] According to further training, the contextualized driving scene description can include oncoming traffic information. This oncoming traffic information enables improved detection and processing of traffic situations, particularly the identification of vehicles approaching from the opposite direction. Oncoming traffic detection can improve the control of the adaptive headlights, as it helps to adjust the vehicle's lighting so as not to dazzle oncoming drivers. This is especially important in situations where the vehicle's cameras and sensors may react more slowly or malfunction, such as in low light conditions, on winding roads, or in adverse weather conditions like fog, snow, or rain.By utilizing a transformer-based image analysis model trained to detect oncoming traffic, the device can overcome these challenges and ensure early and precise dimming of the high beams. Furthermore, the transformer-based image analysis model is able to clearly distinguish reflective objects and non-vehicle-related light sources from actual oncoming traffic, thus avoiding unnecessary dimming and maintaining optimal driver visibility.

[0017] In a training course, the Transformer-based image analysis model can be trained to recognize oncoming traffic based on image data. To train the Transformer-based image analysis model, it can first be fed with a large amount of annotated image data from various traffic scenarios, where relevant features such as oncoming traffic, road layout, traffic density, lighting conditions, weather conditions, and traffic signs are clearly identified. The Transformer-based image analysis model processes the image data in the form of image sections ("patches"), which are embedded into vectors and fed into a Transformer encoder as a sequence along with position information.By employing self-attention mechanisms, the Transformer-based image analysis model learns to grasp spatial and semantic relationships within the scene in order to identify contextual information, such as the presence of oncoming traffic, an upcoming curve, or an urban traffic situation. During training, the Transformer-based image analysis model can be optimized to generate a driving situation classification or semantic description of the scene based on the recognized patterns. This can later be used for advanced functions such as lighting control, driver assistance, or navigation adaptation.

[0018] According to further training, the transformer-based image analysis model can be trained to distinguish oncoming traffic from other reflections or light sources based on image data. For this purpose, a suitably diversified training database can be used during the training of the transformer-based image analysis model. This database can include scenes such as oncoming vehicles with typical headlight characteristics (e.g., position, intensity, distance), reflective traffic signs, guideposts, wet road surfaces, stationary light sources such as streetlights, construction site lighting, or shop windows, and complex weather and lighting conditions (e.g., fog, rain, backlighting). The training data can, for example, be labeled with precise identifiers that differentiate whether a light point originates from a moving vehicle or is a static / reflected light source.

[0019] In advanced training, at least one control signal can be configured to dim the adaptive headlights, particularly the high beams, when indicated based on the contextualized driving scene description. This minimizes the risk of dazzling oncoming traffic due to a failure to dim or a delayed dimming action.

[0020] According to further training, at least one control signal can be configured to keep the adaptive headlights, in particular one of their high beams, undimmed when indicated based on the contextualized driving scene description. This avoids unnecessary dimming.

[0021] In a further development, at least one control signal can be configured to control a segmented front lighting device, in particular its high beam, in a segmented manner. This means that the device is able to divide the front light into different segments and control them independently. This segmented control allows for more precise adjustment of the light beam to selectively illuminate or dim different areas in front of the vehicle. Segmented control of the high beam is particularly advantageous because it enables flexible and dynamic adjustment of the lighting, responding to specific traffic conditions and environmental factors. This can help improve visibility for the driver while simultaneously minimizing glare for other road users.

[0022] According to further training, the electronic processing unit can be configured to provide at least one control signal to or for a vehicle control unit that is designed to control the adaptive headlights. Providing the control signal to the control unit ensures seamless and effective communication between the various vehicle components, resulting in an overall improvement in the functionality of the adaptive headlights.

[0023] According to a second aspect, an adaptive front light is provided for a motor vehicle. The adaptive front light includes a front light device. Furthermore, it includes a device according to the first aspect and its possible further developments.

[0024] The front lighting device can also be understood or referred to as a headlight. This can be, for example, a segmented front lighting device where individual light segments can be activated or deactivated independently. This can be implemented, for instance, as a matrix LED. Other exemplary implementations include pixel light systems with a digital micromirror device (DMD / DLP) for high-resolution light distribution, liquid crystal modulators (LCD shutters) in the light path for dynamic light masking, mechanically adjustable optics (e.g., swiveling lenses or reflectors), and laser-based high-beam systems with adaptive alignment.

[0025] The device according to the first aspect and the front light device can be coupled or connected to each other. The device is designed to control the front light device.

[0026] According to a third aspect, a motor vehicle is provided. This vehicle has a device according to the first aspect. Alternatively or additionally, the motor vehicle has an adaptive front light according to the second aspect.

[0027] The use of the Transformer-based image analysis model enables more precise and faster detection of oncoming vehicles and other relevant objects, leading to timely adjustment of the headlight intensity and thus reducing the risk of dazzling oncoming drivers. Furthermore, thanks to the Transformer's self-awareness mechanisms, the image analysis model can take into account various environmental conditions such as fog, snow, or rain and adjust the lighting control accordingly. Additionally, the model's ability to distinguish between reflective objects and non-automotive light sources and vehicle light sources prevents unnecessary dimming of the headlights, improving visibility and driver safety.

[0028] The invention will now be explained with reference to the figures in the drawings. The figures show: Fig. Figure 1 shows a schematic block diagram of a device for controlling an adaptive front light for a motor vehicle, according to one embodiment. Fig. Figure 2 shows an exemplary motor vehicle with a device for controlling an adaptive front light, according to one embodiment.

[0029] In the figures, the same reference symbols denote identical or functionally equivalent components, unless otherwise stated.

[0030] Fig. Figure 1 shows a schematic block diagram of a device 100 for controlling an adaptive front light 12 for a motor vehicle 10 (see Figure 1). Fig. 2).

[0031] The device 100 comprises an electronic computing unit 110. For example, depending on its architecture, the electronic computing unit 110 can comprise at least one data processing circuit, a CPU (Central Processing Unit), an ECU (Electronic Control Unit), a SoC (System-on-Chip), an NPU (Neural Processing Unit), a GPU (Graphics Processing Unit), a TPU (Tensor Processing Unit), an MCU (Microcontroller Unit), an FPGA (Field Programmable Gate Array), a DPU (Data Processing Unit), and the like. The electronic computing unit can be coupled or connected to the at least one image acquisition device of the vehicle, for example, via a communication network in the vehicle, e.g., CAN, Ethernet, or the like.

[0032] The electronic computing device 110 is configured to receive image data 112 from at least one image acquisition device of the motor vehicle. This image data 112 includes at least one section in front of the motor vehicle 10.

[0033] The image data 112 are fed into a vehicle-based transformer-based image analysis model. This model can be executed, for example, by the electronic computing unit 110. The transformer-based image analysis model is configured to determine a contextualized driving scene description based on the image data 112. Based on this contextualized driving scene description, at least one control signal 114 is generated, which is configured to control the adaptive front light 12, 14.

[0034] The electronic computing unit 110 receives the image data 112 and processes it using the transformer-based image analysis model. The model analyzes the image data and generates a contextualized driving scene description that contains information about the current driving situation, such as oncoming traffic information.

[0035] The control signal 114, generated based on the contextualized driving scene description, is forwarded to the adaptive front light 12, 14. This control signal can dim the front light or leave it undimmed, depending on the detected conditions. In some cases, the control signal can also control a segmented front light device 14 in segments.

[0036] The Device 100 is designed to process image data quickly and precisely to ensure timely adjustment of the headlights. This helps improve visibility for the driver while minimizing glare for oncoming traffic.

[0037] Optionally, the electronic processing unit 110 is configured to provide the control signal 114 to a control unit of the motor vehicle 10, which may also be another control unit, responsible for controlling the adaptive front light. This enables seamless integration of the device into the existing vehicle control system.

[0038] Fig.Figure 2 shows a schematic representation of a motor vehicle 10 equipped with an adaptive front lighting system 12. The adaptive front lighting system 12 comprises a front lighting device 14 and a device 100 for controlling the front light. The device 100 is electronically coupled to the front lighting device 14 to control or regulate its function.

[0039] The vehicle 10 is further equipped with an image acquisition device 16, which serves to capture image data of the vehicle 10's surroundings. This image acquisition device 16 is strategically positioned at the front of the vehicle 10 to ensure an optimal viewing angle for capturing image data. The captured image data is forwarded to the device 100, which is equipped with the aforementioned transformer-based image analysis model. This image analysis model is trained to analyze the image data and generate a contextualized description of the driving scene.

[0040] The contextualized driving scene description includes information about the traffic situation in front of the vehicle, including the detection of oncoming traffic. Based on this description, the device 100 generates at least one control signal 114, which can be used to adjust the light intensity or similar parameters of the adaptive front light 12. In particular, the front light can be dimmed when oncoming traffic is detected to avoid dazzling the oncoming driver.

[0041] Optionally, the device 100 is configured to transmit at least one control signal 114 to a control unit of the motor vehicle 10, which may then take over the actual control of the front light. This enables precise and rapid adjustment of the lighting conditions according to the current traffic situation.

[0042] The front light device 14 may be configured as a segmented controllable device, meaning that it is capable of controlling individual light segments independently. This offers the advantage that specific areas of the light beam can be selectively dimmed or illuminated to ensure optimal visibility and safety. The device 100 is optionally configured to provide such segmented control. Reference symbol list 10 motor vehicle 12 adaptive front lighting system 12, 14 adaptive front lights 14 Front light device 16 Image capture device 100 Device 110 electronic computing equipment 112 image data 114 Control signal QUOTES INCLUDED IN THE DESCRIPTION

[0000] This list of documents cited by the applicant was automatically generated and is included solely for the reader's convenience. The list is not part of the German patent or utility model application. The DPMA accepts no liability for any errors or omissions. Cited patent literature

[0000] CN 119348536 A

[0004]

Claims

[1] Device (100) for controlling an adaptive front light (12) for a motor vehicle (10), comprising an electronic computing device (110) configured to: to receive image data (112) from at least one image acquisition device (16) of the motor vehicle (10), wherein the image data (112) comprise at least one section in front of the motor vehicle (10), to feed the image data (112) into a vehicle-side Transformer-based image analysis model, which is configured to determine a contextualized driving scene description based on the image data (112), and based on the contextualized driving scene description, at least one control signal (114) is to be generated which is configured to control the adaptive front light (12, 14). [2] Device (100) according to claim 1, characterized by , that the contextualized driving scene description includes oncoming traffic information. [3] Device (100) according to claim 1 or 2, characterized by , that the Transformer-based image analysis model is trained to detect oncoming traffic based on the image data. [4] Device (100) according to any one of the preceding claims, characterized by , that the transformer-based image analysis model is trained to distinguish oncoming traffic from other reflections or light sources based on the image data. [5] Device (100) according to any one of the preceding claims, characterized by , that at least one control signal (114) is configured to dim the adaptive front light, in particular one of its high beams, when this is indicated based on the contextualized driving scene description. [6] Device (100) according to any one of the preceding claims, characterized by, that at least one control signal (114) is configured to keep the adaptive front light, in particular one of its high beams, undimmed when indicated based on the contextualized driving scene description. [7] Device (100) according to any one of the preceding claims, characterized by , that at least one control signal (114) is configured to control a segmentally controllable front light device (14), in particular a high beam thereof, in a segmented manner. [8] Device (100) according to any one of the preceding claims, characterized by , that the electronic computing unit is configured to provide at least one control signal (114) to a motor vehicle control unit (10) which is configured to control the adaptive front light. [9] Adaptive front light (12) for a motor vehicle (10), comprising: a front light device (14), and a device (100) according to one of claims 1 to 8, which is coupled to the front light device. [10] Motor vehicle (10), comprising an image capture device (16) and a device (100) according to one of claims 1 to 8 and / or an adaptive front light (12) according to claim 9.

Images